BACKGROUND AND PURPOSE: Functional mobility in people with advanced Parkinson disease, some of whom have a variable response to drug treatment, is often difficult to evaluate. The objectives of this study were to investigate the interrater reliability of measurements obtained with a scale designed to measure mobility and to determine the impact of self-rated dyskinesias and fluctuations on the measure. SSUBJECTS: Twenty-nine people with Parkinson disease and with disability and considerable disease duration (mean=11.7 years, SD=4.9, range=6-22) took part in the study. METHODS: The subjects' performance on a 10-item scale was videotaped. The videotapes were then scored by 2 independent raters, and the scores were used to determine interrater reliability. The stability of 6 repeated measurements was examined in the home situation, taking into account self-rated fluctuations of motor performance. RESULTS: Weighted Kappa values of agreement (.86-.98) confirmed the reliability between testers. Measurement during the "on" phase (when medication was working optimally) and the "off" phase (when the action of medication was strongly decreased or absent) led to different measurements. Measuring frequently within "on" and "off" phases gave relatively stable measurements for total function, bed transfers, and gait akinesia, the latter during the "off" phase only (intraclass correlation coefficients [ICCs]=.70-.93). However, more modest repeatability applied to transfers from a chair (ICC=.65-.67). CONCLUSION AND DISCUSSION: To ensure valid results in future effect studies, clinical differentiation between "on" and "off" phase measurements is proposed on the basis of patients' own perception of their medication status.